Chinese artificial intelligence startup DeepSeek has once again jolted the global AI market by disclosing critical cost and revenue data about its popular V3 and R1 models, claiming a theoretical cost-profit ratio of up to 545% per day.

However, the Hangzhou-based company also cautioned that actual revenue is significantly lower due to several mitigating factors, including free services and variable pricing.

This is the first time DeepSeek has publicly shared any insight into its profit margins from “inference” tasks, a phase in AI deployment where trained models execute tasks such as predictions and chatbot interactions. The company revealed these figures through a GitHub post on Saturday, giving investors and analysts a closer look at the financial dynamics of its models, which have gained global popularity through web and app-based chatbots.

DeepSeek’s Rise As A Cost-Efficient Alternative

DeepSeek’s rise in the AI industry has been nothing short of disruptive. The company first turned heads earlier this year when it revealed that it had spent less than $6 million on chips used to train its models. This is a stark contrast to the billions of dollars that U.S. rivals like OpenAI have invested in cutting-edge hardware. Moreover, DeepSeek relies on Nvidia’s H800 chips, which are significantly less powerful than the hardware deployed by American AI firms.

This development not only questioned the efficiency of U.S. AI firms’ spending strategies but also caused a sell-off in AI stocks outside China. Many investors began to rethink the sustainability of high-cost approaches, especially as DeepSeek’s models, despite using less advanced chips, managed to deliver competitive performance.

The company’s approach has exposed a potential vulnerability in the business models of Western AI firms, which are built on heavy investments in expensive technology. DeepSeek has proven that cost-efficiency can be a viable path to profitability, challenging the notion that only top-tier hardware can support successful AI deployments.

The Numbers Behind DeepSeek’s Model

DeepSeek’s financial snapshot offered a glimpse into its business model, with the rental cost of one Nvidia H800 chip estimated at $2 per hour. According to the company, the total daily inference cost for its V3 and R1 models is $87,072, while the theoretical daily revenue could reach $562,027. If this potential were fully realized, the models could generate just over $200 million in annual revenue, boasting a cost-profit ratio of 545%.

However, DeepSeek was quick to clarify that these numbers represent an ideal scenario. The real-world revenue is substantially lower due to several factors: the lower operational cost of the V3 model, the limited monetization of its services, and discounted pricing during off-peak hours. Furthermore, while some services generate income, many remain free on web and app platforms, limiting profitability.

Censorship Concerns: A Major Roadblock Outside China

Despite its impressive financial model and cost-efficiency, DeepSeek faces a significant barrier to international expansion—censorship. Unlike Western AI models, which are often built on open data sets and trained with a focus on free expression, DeepSeek’s models are required to adhere to strict Chinese censorship laws.

For instance, its chatbots and AI tools are programmed to filter out politically sensitive topics, avoiding discussions on issues like Tiananmen Square, Hong Kong protests, and Taiwan’s sovereignty. This built-in censorship has made the models less attractive to international developers and global enterprises that value unrestricted access to information.

In regions where freedom of speech and openness are crucial—such as the United States, Europe, and parts of Asia—DeepSeek’s censored outputs are seen as a liability, hindering its adoption. Industry analysts have pointed out that developers outside China might be reluctant to integrate DeepSeek’s models into their systems if it means compromising on data freedom.

This censorship issue could impact DeepSeek’s profitability, especially as international markets account for a significant share of AI companies’ revenue. However, DeepSeek has a potential safety net in the Chinese market, which is large and lucrative enough to support sustained growth.

China’s massive domestic market could serve as a buffer for DeepSeek as it navigates international challenges. The country’s booming AI ecosystem, combined with government support for local tech firms, provides a fertile ground for DeepSeek to thrive.

With the Chinese government actively encouraging the development of homegrown technologies, DeepSeek could focus on monetizing its services locally, potentially avoiding the pitfalls of global competition. Moreover, China’s tightly regulated internet space means that censorship compliance might not be as much of a hindrance domestically as it is abroad.

I read comments on my piece on Trump and Zelensky’s show in the White House Oval Office. First, I am not interested if Trump was hard on the Ukraine leader or if the Ukrainian leader was bold, to speak truth to Trump. Those are irrelevant as Ukraine is losing citizens daily and parts of its land remain occupied by Russia.

But here is what I want to focus on: why did Ukraine even need help from the EU, UK and the United States to start with? It needs all the support because of what happened in 2014 when Ukraine toppled its democratically elected president with the support of outsiders. That episode crystallized to where the country is right now.

Go back to the 1990s, that was the scene in Africa. One crazy African would get support from foreigners, cause problems in his nation, and just like that, war begins. Understand that these countries cannot make kitchen knives but they can have supplies of ammunition to fight for years. Who gives them the weapons? Foreign players.

Every African leader or African rebel must learn from Ukraine: no one really cares, and do not allow your country to be used as a vehicle for superpowers to settle issues. Most of the Western leaders who supported the country in 2022 have been voted out of office. And today, it is Ukraine that is working to save its future. But without the signals it got from those world leaders, it would not have toppled its own leaders in 2014!

Finally, morality does not drive geopolitics; only interest does. And as Africans, we must ensure we do not allow those superpowers’ interests to shape our destinies, because when they finish, the victim is the agreeable fellow who accepted to be used!

The Ukraine Maidan 2014, also known as the Revolution of Dignity, was a significant event in Ukraine’s history. It began in November 2013 when then-President Viktor Yanukovych decided to suspend the signing of an association agreement with the European Union, opting instead for closer ties with Russia. This decision sparked widespread protests in Kyiv’s Maidan Nezalezhnosti (Independence Square), which escalated into a larger movement demanding political reform and an end to government corruption.

The protests continued into early 2014, with violent clashes between demonstrators and security forces. The situation reached a critical point in February 2014, when dozens of protesters were killed in confrontations with the police. The unrest ultimately led to the ousting of President Yanukovych, who fled to Russia, and the establishment of an interim government.

The Revolution of Dignity had far-reaching consequences, including the annexation of Crimea by Russia and the ongoing conflict in eastern Ukraine. It also marked a significant shift in Ukraine’s political landscape, with a renewed focus on European integration and democratic reforms.

Google co-founder Sergey Brin has made a rare public move, asking engineers at the tech giant to return to the office five days a week to help improve AI models that could ultimately replicate their own jobs.

The reclusive billionaire, whose net worth is estimated at $144 billion, has personally returned to Google’s Mountain View headquarters, demonstrating his call for a heightened sense of urgency.

The catalyst for this renewed focus on AI is ChatGPT’s launch, which left Google scrambling to regain its footing in a field where it was once a pioneer. Although Google had been at the forefront of AI development, it was OpenAI and its strategic alliance with Microsoft that seized the commercial advantage, putting Google on the defensive.

In a memo seen by The New York Times, Brin wrote to engineers working on Google’s Gemini AI models, stressing that the “final race to AGI (Artificial General Intelligence) is afoot”. He expressed confidence that Google had “all the ingredients to win this race”, but emphasized the need to “turbocharge” efforts. His prescription for success: “60 hours a week is the sweet spot of productivity.”

Brin also encouraged engineers to use Google’s own AI models to write their code, arguing that doing so would make them “the most efficient coders and A.I. scientists in the world.” This directive aligns with a broader trend where tech leaders are promoting AI tools as a means to enhance productivity, but it also exposes a deeper irony: Brin is effectively asking engineers to use the same technology that might eventually make their roles redundant.

The Irony of AI-Driven Efficiency

Generative AI, such as Google’s Gemini, works by ingesting large amounts of data and recognizing patterns to generate new content, including code. In theory, this technology could automate a significant portion of coding tasks, leading to higher efficiency. Other tech leaders, like Salesforce CEO Marc Benioff, have already indicated that AI agents have advanced to a point where they are reducing the need for human engineers. Benioff stated during an earnings call that Salesforce would not be hiring more engineers this year, attributing this decision to the success of AI in handling tasks previously managed by human staff.

However, it is important to view such statements with skepticism. While AI advocates highlight its potential to cut costs and improve productivity, many believe that company leaders might be using the hype around AI as a pretext to reduce headcounts, save on labor costs, and appease investors. For instance, Salesforce had earlier cut 10% of its workforce—about 7,000 employees—under pressure from activist investors to improve profit margins.

AI’s Limitations: Code is Not Just Code

Though AI tools can automate boilerplate coding, they struggle with complex, large-scale codebases due to memory constraints. Additionally, while AI can generate code snippets, engineers need to understand the underlying logic to fix bugs and implement improvements. Ironically, companies like Anthropic, a prominent player in AI safety research, explicitly ask job applicants to certify that they will not use AI during the application process, highlighting the limitations of AI-generated work.

The fear among engineers is not just that AI might replace them, but that companies may choose to use AI even if it performs worse than humans, purely as a cost-saving measure. This dynamic is reminiscent of a scenario where a manager asks a senior employee to train their younger, cheaper replacement.

Proponents vs. Skeptics: Two Sides of the AI Debate

Proponents of AI argue that the technology will lead to more work, not less, by freeing up engineers to focus on more complex projects. By automating mundane coding tasks, engineers could theoretically build more products and achieve greater innovation. However, skeptics believe the push for AI adoption is less about empowering engineers and more about reducing costs and streamlining operations.

The debate extends beyond productivity to the broader dynamics of workplace control. The return-to-office mandate is not just a Google phenomenon but part of a wider trend among corporate executives seeking to reassert authority over workers who gained greater flexibility during the COVID-19 pandemic.

Power Shift in Silicon Valley

The tech industry, particularly Silicon Valley, has seen a power shift. Engineers once highly sought after and empowered by remote work opportunities, now face reduced leverage as companies like Google reverse their remote work policies. This shift comes amid a backdrop of mass layoffs and a tightening job market, which has allowed companies to demand more from remaining staff.

Tech giants, including Google, are also incentivized to bring employees back to the office to justify the billions of dollars spent on lavish headquarters. For example, Google’s Mountain View campus, with its futuristic architecture and state-of-the-art amenities, represents a significant investment that the company would prefer not to waste.

Brin’s memo adds an ember to the AI debate. On one hand, his call to arms reflects a genuine urgency in a high-stakes competition with OpenAI and Microsoft. On the other, it highlights a paradox in the AI industry: engineers are being asked to build the very tools that might render them obsolete.